Clinical Translation of an Insertion Tool for Minimally Invasive Cochlear Implant Surgery

Cochlear implants (CIs) have helped restore the perception of sound to hundreds of thousands of individuals worldwide [1]. Traditional CI surgery requires a mastoidectomy to access the inner ear structures. Mastoidectomy involves removing a pocket of bone from behind the external ear (see Fig. 1(a)), a procedure during which the surgeon must rely on experience, hand-eye coordination, tactile feedback, and visual cues to avoid vital structures [2]. These vital structures include the facial nerve, which controls motion of the face, and the chorda tympani, which controls some components of taste. There have been a variety of efforts to minimize the invasiveness of this procedure and obviate the need for mastoidectomy. Some of these approaches include the pericanal technique, the transcanal technique, and the suprameatal approach [3–8]. These methods have successfully avoided mastoidectomy; however, the angle of approach for insertion of the electrode array (EA) into the cochlea is suboptimal [9].

Recent research has investigated whether image-guided approaches that leverage patient-specific anatomical knowledge could be used to avoid mastoidectomy in favor of a single drilled linear trajectory to the cochlea, termed minimally invasive cochlear implant surgery (MICIS) for the remainder of this paper (see Fig. 1(b)). Specifically, this method calls for the creation of a drilled path to the channel within the cochlea closest to the auditory nerve endings, the scala tympani (ST). A variety of approaches have been employed to guide the drill along the specified trajectory including a parallel robot [10,11], an industrial robot arm guided by optical tracking of the robot and the patient [12,13], a hand drilled conical path using image guidance [14], a custom serial robot arm in conjunction with optical tracking [15], and use of microstereotactic frames [16].

All of these approaches share a common theme in that they help to realize a minimally invasive single drill path to the cochlea which avoids critical anatomy. However, the surgically created tunnel creates geometric constraints for the surgeon to insert the active end of the CI—the EA—into the cochlea. The surgeon is tasked with threading the EA down a narrow tunnel of spiculated bone, punctuated with irregularly shaped air cells, through the middle ear air gap, and into the ST with highly restricted visual feedback. During insertion, the EA can sometimes blindly detour into a partially opened mastoid air cell and/or buckle in the middle ear space. This deflection is a particular issue for the more flexible straight EAs (as opposed to the precurved EAs that have a stylet resulting in a stiffer overall structure).

As a result, despite promising results and potential advantages of a single tunnel approach, i.e., it is less invasive, safer for vital anatomy, and facilitates an optimal insertion vector—the final step of EA insertion through the tunnel and into the cochlea remains a challenge. Figure 2 displays the approximate shape of the surgical space resulting from MICIS [16]. In this work, a solution to this challenge using a novel manual insertion tool for MICIS is described.

A variety of automated insertion tools for CI EAs have been developed in the past featuring innovative designs [17–24]. These tools are capable of achieving super-human accuracy and speed. While these automated solutions could prove very useful in the future, the goal of this work was to develop a simple manual tool that could be quickly translated to the operating room. Manual insertion tools include traditional surgical instruments (e.g., forceps, claws, and guidance tools such as Ref. [25]), which are not designed specifically for MICIS. Typically, these tools are not compatible with the MICIS workspace, and if they are, they do not solve the inherent challenges associated with this approach. Specific manual MICIS tools include a tool by Caversaccio et al. who used a metal split-tube catheter design [26], but this tool is not compatible with the workspace of Ref. [16] to solve the aforementioned challenges. Kratchman et al. developed a tool [27] compatible with the workspace of Ref. [16]; however, this tool is not compatible with larger diameter, straight EAs that are very susceptible to buckling [28].

In summary, user requirements for this tool are that it be a manual (i.e., not automated) tool that enables a clinician to perform the device's intended use which is to thread a straight EA down the porous tunnel of bone and insert it safely and effectively into the cochlea. These requirements lead to the following design requirements that the tool be (i) manually actuated, (ii) able to accommodate the MED-EL (MED-EL, Innsbruck, Austria) Standard and FLEX series EAs, (iii) able to accommodate any viable MICIS candidate, (iv) sterilizable and able to be fully disassembled from the EA, (v) must not damage the EA, (vi) must demonstrate an improvement over traditionally used tools, and (vii) it must not limit the final EA insertion depth.

Recently, a simple, disposable manual insertion tool was described [28,29] that can insert large diameter EAs such as the MED-EL Standard EA (nominal diameter of 1.3 mm, maximum diameter of 1.4 mm) in MICIS. Printed on a low-cost three-dimensional (3D) printer, it utilizes a split cylinder design with roller wheels slightly compressing and advancing a CI EA through a distal polyimide sheath that spans the middle ear space and guides the EA into the cochlea. The roller tool can be activated blindly, enabling a surgeon to visualize the middle ear and cochlea opening while the EA is advanced.

Prior work [28,29] focused on initial concept design and feasibility in phantoms and cadavers. This paper focuses on refinement of the design and the steps needed to achieve U.S. Food and Drug Administration (FDA) investigational device exemption approval to enable the continuation of MICIS clinical trials. Specifically, this work fulfilled specification of design inputs, outputs, verification, and validation. The contributions of this paper are as follows: (1) population analysis (N = 97) of patient-specific parameters for both roller tool design and phantom model generation and verification that patient-specific designs can be suitably built for all viable candidates, (2) developing a method for and testing roller tool disassembly around an EA, (3) improving tool concentricity, (4) miniaturizing the roller tool for compatibility with the geometric requirements of the rest of the CI system (i.e., subcutaneous receiver permanently attached to the EA), (5) creating a clinically realistic MICIS testing setup that could be used as a general surgical trainer for MICIS in the future, (6) evaluating the insertion force profile of the roller tool versus surgical forceps, and (7) evaluating cadaver performance with full mock implantation including disassembly and removal of the roller tool, with all work leading to (8) clinical translation in a first live human case.

In this section, methodology toward contribution 1 to determine the MICIS population parameters (i.e., distances along the drill path of the air gap, medial drilling, and lateral drilling in Fig. 2) for the implementation in Ref. [16] is described. These distances were quantified for three main reasons. First, such analysis would help determine if the insertion tool could be a one-size-fits-all design or if the 3D printed housing and/or distal polyimide sheath length would have to be adjusted on an individual basis. Second, population analysis would enable the design of a minimally invasive phantom model representative of the surgical population. Finally, this analysis could provide further insight into anatomic nuances that may rule out some candidate patients from MICIS.

Ninety-seven temporal bone scans were analyzed under IRB-approved protocols. All scans were acquired on clinical CT scanners, and median voxel size for this dataset was 0.25 mm × 0.25 mm in the plane of scan and 0.34 mm between slices. For each patient CT scan, MICIS was planned, and the resulting drill path was analyzed. To perform this planning, the temporal bone anatomy was first segmented using an atlas-based approach [30–32]. A third-year otolaryngology resident surgeon verified that all segmentations were anatomically valid. Any questionable segmentations were further validated by a very experienced CI surgeon (>500 implants over 20 years).

Based on the segmented anatomy, a drill trajectory was generated automatically [33]. Critical points along the trajectory (t, m, a, l, and s) were identified as shown in Fig. 2. All points lie along the trajectory axis. Point t (target point) is the most lateral (toward skull surface) point of the ST along the axis. Point m is the end of the medial drilling, selected to lie within the middle ear space and at least 1 mm past any mastoid bone given that the medial drill tip is conical with a height of approximately 0.8 mm. Thus, this selection ensures that the medial drill bit enters the middle ear to its full diameter. Next, the first point along the trajectory moving laterally from the cochlea where the drill path was surrounded by 50% of bone was identified and labeled a, specifying the boundary of the middle ear air gap. The level/window for determining this point was set to 500/1000 HU for all scans for standardization; interscanner intensity variability was neglected.

Point l was chosen 10 mm behind point m based on established drilling protocols with the medial drill bit length being either 10 mm or 13 mm longer than the lateral drill bit length. The shorter differential is preferred given better access and more stable drilling. If this l (10 mm from m) placed the lateral drill bit within 4 mm of the facial nerve or within 0.4 mm of the external auditory canal (EAC), l was placed 13 mm lateral to m. If this trajectory still resulted in the lateral drill path being closer than 4 mm to the facial nerve or 0.4 mm to the EAC, the candidate was omitted from subsequent analysis and was not considered a MICIS candidate. Finally, intersection of the skull surface with the midline trajectory was labeled as point s.

This section briefly describes the basic tool design concept originally presented in Refs. [28] and [29]. The new tool that is the subject of this paper was created by integrating lessons learned during the clinical translation process, as stated in contributions 2–4. These contributions are the development of a tool disassembly method, improving device concentricity, and overall tool miniaturization. First, full removal of the EA from the tool is now realized using slit polyimide. Disassembly proceeds by holding the EA stationary with tweezers and first removing the tool half without the polyimide attached. Then, while still holding the EA in place, the polyimide-attached tool half is pulled out via the slit. Of note, the polyimide has a helical slit with a high pitch (see Fig. 3(a)), which enables it to be removed around the transmitting wire while minimizing the chance that the EA will shift out of the slit during insertion. A second tool design change is the addition of the second tongue-and-groove feature on the guide cylinder of the tool (see Fig. 3(a)) to ensure the two tool halves remain concentric with one another. Finally, miniaturization of the tool was necessary for clinical use because the EA is permanently attached to an internal receiver buried below the soft tissue of the skull behind the ear (see Fig. 4). The results from Sec. 2.1 were used to advise the miniaturization (i.e., dimensions r, h, and g in Fig. 3(a)) as described in Sec. 3.2.

In general, this insertion tool design is composed of two 3D printed halves that slide together with a tongue-and-groove joint. Each tool half has a roller wheel to aid in EA advancement. A polyimide sheath is glued to the end of one tool half with a drop of medical grade superglue. This tool was designed to accommodate the MED-EL Standard and FLEX series EAs but could easily be modified to accommodate other EAs. The MED-EL Standard EA has a silicone diameter of approximately 1.3 mm around the transmitting wires (as do the FLEX series EAs), but the Standard has a 1.4 mm silicone stopper ring indicating full insertion. For the remainder of this paper, the electrode transmitting wires housed in silicone will be referred to as “the transmitting wire” for simplicity. To use this tool, the EA is threaded into the lumen of the tool and once the last electrode is past the wheels, the wheels contact the transmitting wire and can be used to advance the EA slowly and consistently out of the tool and into the cochlea.

Device assembly was standardized using the assembly jig shown in Fig. 3(b). Most of the tool (mandrel, wheels, axles, and tool halves) is made of Dental SG Resin on a Form 2 (Formlabs, Somerville, MA) 3D printer. This material adheres to EN-ISO 10993-1:2009/AC:2010, EN-ISO 20795-1:2013, and EN-ISO 7405:2009/A1:201 standards. A drop of Loctite 4981 superglue that adheres to ISO-10993 standards (Henkel AG & Company, Düsseldorf, Germany) holds the distal polyimide sheath (Microlumen^®, Oldsmar, FL) to one of the paired entry chambers. The polyimide sheath is chemically inert and adheres to requirements for U.S. Pharmacopeia Class VI compliance. The tool is 3D printed to enable design flexibility and to increase the speed and ease of fabrication. The Form 2 3D printer has a maximum resolution of 0.025 mm and a spot size of 0.14 mm, so this manufacturing method is suitable for standard machining tolerances. All 3D printed components were printed according to the protocols for medical grade devices available on the Formlabs website. (This material is used regularly for creation of dental drill guides for dental implants). According to a Formlabs study involving 84 printed Dental SG surgical guides, 93% of the intaglio surface area was accurate within ±0.1 mm, demonstrating high accuracy and repeatability [34]. Sterilization validation of the fully assembled tool (Fig. 3(c)) was performed at SMP GmbH (Tübingen, Germany), and the device was validated to be wrapped and sterilized at 132 °C for 4 min.

In this section, contributions 5 and 6 are described: development of a MICIS phantom and evaluation of the insertion force profile noting that higher forces have been correlated with increased cochlear trauma and poorer patient hearing outcomes [36–38]. The experimental setup (Fig. 5) was designed to mimic that of an actual MICIS using the scan analysis described in Sec. 2.1 and the determined mean values. This model differs from a previous minimally invasive phantom model [39] in that it is designed for clinically realistic insertion force profile evaluation instead of for drilling accuracy evaluation. The described model could be used as a general surgical trainer for MICIS, as practice for a specific MICIS candidate, or to compare insertion methods.

To create a clinically realistic model, the patient scan with the closest parameter values to the mean values determined in the analysis (||t − a|| = 5.3, ||a − l|| = 8.0, ||l − s|| = 14.4) mm (see Sec. 3.1) was selected. The values of the selected patient were (5.5, 7.5, 14.1) mm. The segmented ST object file was exported, and the skull was isosurfaced. Meshmixer (Autodesk, San Rafael, CA) was used to trim the skull object to a reasonable size that still included all vital bone for the procedure visualization and insertion. Meshmixer was also used to modify the bone to be 3D printable, i.e., fill holes and extrude the skull surface to generate a 3D printable mesh. After generating the printable object file in Meshmixer, Solidworks (Solidworks, Waltham, MA) was used to assemble all exported anatomy with their origins coincident and coordinate systems aligned such that they were in the correct anatomical orientation relative to each other. The remainder of the assembly was designed to accommodate and mount the temporal bone and a force transducer (Nano17, ATI Industrial Automation, Apex, NC; SI-12-0.12 calibration). With this calibration, the transducer has a sensing range of 12 N in the X–Y axes, 17 N in the Z axis, and a resolution of 1/320 N in all axes. Data were acquired using the ATI Network Force/Torque sensor system. Note the decoupling gap in Fig. 5 that was incorporated to decouple forces on the temporal bone from forces on the ST. The ST model and ST holder were both printed using stereolithography. The remainder of the components were printed using fused deposition modeling.

Insertion force profile data were collected to demonstrate phantom model functionality and to ensure that the tool was not introducing any undesirable force spikes or artifacts as compared to standard forceps insertion. IRB approval was acquired for this user study, and an expert surgeon inserted an unused MED-EL Standard EA into the soapy-water-filled ST model four times using the roller tool and four times using traditional tools (i.e., surgical forceps), with the order chosen randomly for each of the four comparisons. Force data (1 kHz) and video (60 fps) of the ST were recorded for each trial (experimental setup shown in Fig. 5(c)), with the surgeon blind to these recordings. Individual and average insertion force profiles versus angular insertion depth are reported in Sec. 3.3.

In this section, the performed cadaver study for simulated-use human factors validation testing to evaluate clinical viability of this device is described (contribution 7). Cadaver insertions were performed primarily to verify that (1) the tool dimensions accommodate receiver placement, (2) the removal method is effective, and (3) the EA is unharmed during insertion. An unused MED-EL Standard EA was used for each of the eight ears studied. A MED-EL Synchrony template receiver was attached to each EA at the same distance as a clinical CI (see Fig. 4). Pictures were taken of the EA before and after insertion to document any potential damage. Saline was used as a lubricant for all eight ears. For the final head, soapy water was also used for each ear (after saline insertions had been performed).

Minimally invasive cochlear implant surgery was performed as previously described [16] and is summarized in Fig. 6. First, a pre-operative scan was taken using the Xoran XCAT™ CT scanner (Xoran Technologies LLC, Ann Arbor, MI). Critical anatomy was segmented and a safe drill path was automatically planned as described in Sec. 2.1 (Fig. 6(a)). An expert surgeon validated the segmentations for clinical viability. After segmentation validation, the tool was adjusted to be patient-specific—the polyimide sheath was cut to length based on the patient-specific drill path. The transmitting wire was marked at the point at which insertion would be complete relative to the most lateral point of the insertion tool (see Fig. 4) (i.e., when this mark was about to enter the insertion tool, the EA was fully inserted within the cochlea). The tool was then sterilized. While the tool was being sterilized, fiducial anchors were placed (Fig. 6(b)), and an intra-operative scan was taken. This intra-operative CT scan was registered to the pre-operative CT scan using a standard, intensity-based mutual information method [40,41] with Vanderbilt University Cochlear Implant Planning software. Such methods have been shown in multiple independent studies to lead to subvoxel registration errors [42]. The Microtable^® microstereotactic frame was designed using automatic custom matlab (Mathworks, Natick, MA) code, and the patient-specific Microtable was fabricated, assembled (Fig. 6(c)), and verified with postfabrication measurements. During Microtable assembly, the surgeon lifted the tympanomeatal flap for access to the middle ear, created a small periosteal incision for the drill path, and created a subperiosteal pocket for the receiver, as in standard CI surgery.

The Microtable was mounted, lateral drilling (3.8 mm diameter) was performed (Fig. 6(d)), and a verification CT scan was taken. After visual verification of the path, medial drilling (1.6 mm diameter) was performed, and the Microtable was removed (Fig. 6(e)). The round window overhang was taken down with a 1 mm diamond drill bit, and the round window membrane reflected to gain ST access.

For each ear, an insertion using an unused MED-EL Standard EA was done with traditional tools (i.e., forceps) and with the roller tool (Fig. 7(f)). A CT scan was acquired after insertion was deemed complete. For roller tool insertions, the tool was then fully removed (Fig. 7(g)), and another CT scan was acquired. The ST was refilled with saline between each insertion, and the tool and EA were dipped in distilled water between each insertion. For the final head, soapy water was also used to test full insertion capabilities of the tool.

After the rigorous testing reported above, the updated roller tool and the steps described in Fig. 7 were incorporated into the pre-existing workflow (Fig. 6) for MICIS clinical trials at Vanderbilt (contribution 8). Approval of an amendment to a previously submitted investigational device exemption from the FDA was obtained prior to clinical use allowing use of the roller tool as part of this feasibility trial.

In the analysis, 19 of the 97 patients (20%) would not tolerate a MICIS plan given proximity limitations most typically with the lateral aspect of the EAC. This result is fairly consistent with that reported in Ref. [43] for a drill bit diameter of 1.6 mm where exclusion was determined based on facial recess size and system accuracy instead of absolute proximity to adjacent structures. Interestingly, most of these omissions needed to be determined visually on the CT scan and not automatically using only the segmentations. Automated proximity analysis between structures only identified 4 out of the 19 of the candidates for omission because it was the lateral-most portion of the EAC that was the violating area most often—a region of the EAC not typically segmented. Of the remaining 78 patients who could accommodate a MICIS trajectory, 18 required the 13 mm differential between the medial and lateral drill bits (i.e., the larger diameter, lateral drill bit did not go as far into the skull). As with those patients excluded from MICIS, the primary reason for the larger differential was proximity of the lateral drill to the EAC. In the automated proximity analysis, 29% of plans collided with the segmented chorda tympani, a percentage slightly higher, but fairly consistent with past reporting [33,44]. It is also important to note that 30 ears out of the 97 ear dataset came from 15 heads. In other words, the left ear and right ear were both included from those 15 heads. Ears from the same head were not excluded because there was no evidence that the whole MICIS surgical workspace will behave symmetrically. In three cases, one ear of a head met the omission criteria, and the other did not.

After this analysis, histograms were generated for the parameters: ||t − a||, ||a − l||, and ||l − s|| with N = 78 after omissions (Fig. 8(a)). The hypothesis of normality for each parameter could not be rejected by a Kolmogorov–Smirnov test with 5% significance (p = 0.96, 0.08, 0.50), respectively. The mean for each parameter is plotted as a horizontal line in Fig. 8(b). The mean values are (5.3, 8.0, and 14.4) mm, respectively. The overall drill path lengths (addition of the three distances in Fig. 8(b)) were between 21 mm and 36 mm with a mean of 28 mm, consistent with past reported approximations of 30 mm [13,45] and maximum of 35 mm [10].

Before proceeding, it is important to note how repeatable these study results are since the automatic drill trajectory generated is not an analytic result but rather a probabilistic result with many possible solutions that is generated by finding an optimal path that does not violate anatomic restrictions. To investigate repeatability of such trajectories, N = 100 independent drill trajectories were generated automatically by wrapping the methods detailed in Ref. [33] in a loop and applying the automated algorithm 100 times for each patient scan. For 94/97 scans, this method was successful at generating 100 trajectories. For the remaining three scans, only a subset of 100 acceptable trajectories could be found. For a given MICIS candidate, this enabled analysis of how much the drill path entry point and target point would vary if the path were regenerated using the automated algorithm. Furthermore, by looking at the median maximum variation for the population, insight is gained into study repeatability.

For these independent drill paths for each scan, the maximum deviation of the target point, entry point, and path length was analyzed. The median maximum deviation for the target point was 0.51 mm, the median maximum deviation for the entry point was 0.92 mm, and the median maximum path length deviation was 0.07 mm. Figure 9 shows a plot of the median deviation of the entry point on the horizontal axis and median deviation of the target point on the vertical axis of an ellipse drawn for each of the 94 scans with N = 100 trajectories, where outliers can be viewed by those ellipses with low opacity. Note—the path length in this simulation is only accounting for the distance between the automatically generated target point to the automatically generated entry point in the facial recess. The median maximum values show that a given patient's optimal insertion vector parameter lengths tend to have small variation, supporting the idea that results from the parameter analysis would be repeatable. Parameter analysis results were used to update the tool design and inform phantom model creation as will be described next.

The parameter analysis indicated that the same 3D printed tool could be used for all patients (parameters h, r, and g in Fig. 3(a)), but the distal tube would need to be customized to account for interpatient variation. Fortunately, the distal tube is a flexible and easily trimmable sheath made of polyimide. In regard to the 3D printed tool dimensions, initial cadaver studies with disassembly revealed that the slack in the EA transmitting wire once the receiver is placed was a limiting factor in tool geometry above the skull surface. This constraint is highly dependent upon the distance between the drilled tunnel and internal receiver placement which is variable due to patient and surgeon preference. Considering an EA fully inserted in the cochlea, the distance from the cochlea opening to the internal receiver center measures approximately 121 mm [35]. Subtracting off the maximum drill path length (from Sec. 3.1 = 36 mm) leaves 85 mm outside of the skull. The distance from the drill tunnel opening on the surface of the skull to the center of the subcutaneously placed internal receiver will be approximately 69 mm (79 mm average from Ref. [46] minus an assumed 10 mm distance between the EAC and the drill tunnel opening). Given these restrictions, the sum of the tool height (h) and tool radius (r)—that are external to the skull—needs to be less than approximately 16 mm. The dimensions h and r were specified to meet this constraint and be the minimum possible that still enables wheel actuation, which was found to be h = 7 mm and r = 4.5 mm. The sum of this choice of h and r is well below the constraint to give the surgeon the most flexibility in placement of the internal receiver and also take into consideration the curvature of the surface of the skull, which adds between 0.6 mm and 4.5 mm to the overall path length for the tool radius, with a mean of 2.8 mm.

Based on the parameter analysis, the 3D printed guide cylinder of the tool was made 14 mm long, which is the lower bound of the 95% confidence interval of ||l − s|| (rounded to the nearest whole number for ease of verifying distances in the operating room) to work for the largest number of patients possible. If the mean or upper bound was chosen, the surgeon may risk having too much of the tool outside of the skull, without enough slack to insert the EA through the tool. To verify this choice, Fig. 10 plots the estimated wire slack for a range of choices of guide cylinder lengths (g) and all viable candidate guide cylinder spaces (||l − s|| + skull surface addition for each scan). The choice of g should be based on maximizing wire slack while minimizing the amount of the EA in the polyimide tube when possible to reduce friction. Guide cylinder space is plotted on the x-axis (range [9.4, 24.1]), possible choices of g are plotted on the y-axis (range [5, 25]), and the resulting estimated wire slack for all 78 viable candidates is plotted on the z-axis (range [−1.6, 20]). This plot demonstrates that the 14 mm choice occurs at approximately the knee of the graph (plane drawn at y = 14 mm) and was determined to be a good tradeoff between maximizing slack for a large number of patients and minimizing friction.

Results from the phantom user study are shown in Fig. 11. Force data were binned into 5 deg increments. For standardization, an insertion was deemed complete after the force exceeded 34 mN, the minimum highest force for any of the eight trials. Additionally, forces that corresponded to 0 deg of movement were omitted from analysis because the stop and start of movement throughout each manual insertion was not consistent between trials. Since data are binned to generate an average insertion profile, this stationary force data would skew the results (e.g., lead to inappropriate weighting of low forces if the EA was left in place while readjusting the microscope in one trial and not another). Thick lines indicate the binned average of the roller tool insertions (dashed line) and forceps insertions (solid line), smoothed using a moving average over 5 points (i.e., over 25 deg). Figure 11 shows similar insertion force magnitudes for both insertion techniques. Roller tool insertions exhibited slightly higher forces but more consistently deep angular depths (note the forceps trial that only reached about 180 deg). More importantly, the roller tool did not introduce any unexpected large force spikes (all roller insertions reached at least 310 deg before exceeding the 34 mN threshold).

Results from the cadaver study (N = 8 ears) demonstrated that the roller tool qualitatively increased the ease of insertion as reported by the surgeon. Postexperiment EA pictures indicated that the positions of the individual transmitting wires in the transmitting wire bundle and the individual electrode contacts on the EA were qualitatively unchanged by insertions through the roller tool. Tool sterilization, sizing, and removal were successful in all cases. Quantitatively, insertions with the roller tool resulted in angular insertion depths that were on average 15% deeper than insertions using the traditional tool. Using a one-sided paired t-test analysis, this percent increase was statistically significant for a significance of 10% (p-value = 0.07).

The average percent difference in angular insertion depth between when the tool was still in place and when it was disassembled and removed around the EA was 5%. Using a two-sided paired t-test, this percent difference was not statistically significant for a significance of 10% (p-value = 0.65), indicating that tool removal did not have a statistically significant effect on final EA position. In interpreting these results, we note that the sample size is relatively small, and additional experiments would be useful to improve the robustness of these statistics in the future.

For both ears of head four, after testing with saline as an intracochlear fluid on each side, soapy water was used as a lubricant to decrease friction further and ensure that the tool would not restrict the EA from reaching full insertion for any reason. In the left ear, the resulting angular insertion depths were 551 deg, 696 deg, and 682 deg for the forceps, roller tool after insertion, and roller tool after tool removal, respectively. In the right ear, the corresponding angular insertion depths were 424 deg, 536 deg, and 545 deg. These results validated that the chosen tool dimensions did not obstruct the EA from achieving a full insertion in a cadaveric cochlea with the subcutaneous receiver in place.

After rigorous preclinical testing, this tool was used in an investigational human feasibility trial of MICIS for EA insertion. The tool was sterilized per the sterilization protocol described earlier and delivered to the operating room. After drilling was complete, steps (f) and (g) of Fig. 7 were performed, and the roller tool enabled threading of a MED-EL FLEX²⁸ EA down the drilled tunnel, through the middle ear space, and into the cochlea after the internal receiver was placed. The tool enabled a full insertion of the EA, with all 12 electrodes intracochlear in the ST and a final angular insertion depth—after disassembly and removal of the tool—of 557 deg.

Through design verification and validation, user and design requirements for the tool have been demonstrated to be met. Requirements (i) and (ii) were verified with initial design and testing presented in prior work. In Ref. [29], it was shown that the new manual tool design would enable insertion of MED-EL Standard and FLEX series EAs, and in Ref. [28], this result was further validated with cadaveric workflow, but without full tool disassembly or receiver placement. In this work, a population parameter analysis was performed to determine whether a single 3D printed tool could accommodate all viable candidates, and this analysis was used to specify tool dimensions to accomplish requirement (iii). This population analysis indicated that one 3D printed tool could be used with polyimide adjustments to make the tool patient-specific. In this analysis, approximately 1 in 5 CI candidates have anatomy that is not favorable for MICIS. The determined results of the distances ||t − a|| and ||t − s|| are generalizable to any MICIS with a similar choice of insertion vector (i.e., tangential to the basal turn). While other parameters may be more specific to the Vanderbilt implementation of MICIS, this analysis shows that the middle ear (||t − a||) and facial recess (||a − l||) depths have less variability than the lateral mastoid depth (||l − s||). Additionally, the analysis highlighted that EAC proximity to the lateral drill path should be visually checked on the surgical plan to ensure patient safety (15/19 omitted candidates were excluded based on this visual check in this study). Automatic drill path variability investigation showed that, for a given patient, path variation tended to be small. This is an important result given that drill path determination is a probabilistic and not an analytic result. Small variation implies that results from this study should be relatively replicable.

Next, for design requirement (iv), an external company validated a sterilization protocol for the tool since the standard sterilization routines for the Dental SG resin from Formlabs could not be guaranteed to be valid with the addition of the polyimide and superglue. A complete disassembly method was developed that featured helically slit polyimide and resulted in very little change to the final angular insertion depth (approximately 5%). Avoidance of EA damage (requirement (v)) was qualitatively determined using high resolution pictures of the EA before insertion, after insertion through the tool, and after insertion into a cadaveric cochlea. These pictures indicated that tool use likely will not damage the transmitting wire or active end of the EA.

For each experimental evaluation, the roller tool was compared to surgical forceps for fulfillment of design requirement (vi). Before introducing a new tool into the operating room, it is important that both the device developers and the clinicians/surgeons are confident that the new tool introduces a significant advantage over traditional tools to warrant its introduction. Experimental evaluation included a comparison of the insertion force profiles where it was demonstrated that the roller tool did not introduce any undesirable force spikes and demonstrated similar force magnitudes to the surgical forceps. The described phantom model was successful for force profile evaluation on a clinically representative MICIS model. Furthermore, this test setup could be used as a general surgical trainer in the future for surgeons learning MICIS EA insertion, for surgeons practicing for a particular MICIS candidate, or to compare different insertion approaches on the same model as was done in this work. A similar model could be created using the details described in Sec. 2.3 for a different MICIS candidate.

In cadaveric insertions, the roller tool yielded 15% higher insertion depths when compared with forceps. Additionally, it was qualitatively demonstrated in both phantom and cadaver insertions that the roller tool made it significantly easier to thread the EA into the cochlea opening (a nontrivial task when using the forceps due to visualization and dexterity limitations, as well as the prevalence of buckling of the EA in the middle ear space). All of these results indicated that the roller tool introduced a sufficient enough advantage to move it to the operating room. Finally it was verified that the chosen tool dimensions based on the population analysis did not obstruct the EA from achieving a full insertion in a cadaveric cochlea with the subcutaneous receiver in place (requirement (vii)). This finding is important because insertion depth has been correlated with postoperative audiologic performance [47] and thus, should not be restricted whenever possible.

With the user and design requirements met, this tool was translated to the operating room for the first time, and enabled a full ST insertion in a clinical case [48]. No backup tools were needed, although three were available. There are many important areas of future work for this tool. The tool should be evaluated with different straight EA types (simply by modifying the inner diameter and bottleneck diameter of the tool) to verify functionality across types. Additionally, future work could investigate the addition of a stylet arrestor to enable insertion of precurved EAs with this tool, although the buckling issue is much less prevalent with these stiffer EAs in MICIS. A third area of future work should be expanding the number of clinical cases to include more patients and achieve a statistically significant measure of clinical success. Another interesting study would be evaluating whether this tool could be useful for traditional, mastoidectomy-based CI surgeries as opposed to only the minimally invasive situation. A final area of future work would be determining how beneficial this tool is to novice users (in MICIS and in traditional CI surgery) compared to its usefulness to expert users in a similar setting.

The developed insertion tool outperforms traditional tools (i.e., surgical forceps) used for insertion of EAs in MICIS qualitatively by ease of use and quantitatively by angular insertion depth. In its current design, this tool can only be used with straight, nonstyleted EAs. Successful introduction into the operating theater was performed for the first time.

The authors thank Anandhan Dhanasingh of MED-EL for the EAs used in experiments. R.F.L. is a consultant for Advanced Bionics and Spiral Therapeutics. This project was supported by Award No. R01DC008408 from the National Institute on Deafness and Other Communication Disorders of the National Institutes of Health and by the National Science Foundation Graduate Research Fellowship No. DGE-1445197/1937963.

• National Institutes of Health (No. R01DC008408; Funder ID: 10.13039/100000002).

• National Science Foundation (No. DGE 1445197/1937963; Funder ID: 10.13039/100000001).